Real-Hardware Deployment of a Nussbaum-Function PID Controller on a Current-Controlled Low-Cost Actuator via...

Area of Science:

• Robotics
• Control Systems Engineering

Background:

• Real-world deployment of adaptive controllers faces challenges due to hardware limitations like friction and latency.
• Existing controller formulations often fail to account for these real-world dynamics.

Purpose of the Study:

• To bridge the gap between simulation-level adaptive controller design and real-hardware deployment.
• To investigate and improve a Nussbaum-function PID controller for a low-cost robotic manipulator.

Main Methods:

• Translated a Nussbaum-function PID controller from simulation to direct current-command control on a Niryo NED3 Pro actuator.
• Isolated single-actuator dynamics to focus on deployment-layer behaviors.
• Developed a hardware-oriented implementation with adaptation-state regularization, velocity-reference feedforward, and tail-region damping.
• Optimized controller parameters using a hardware-aware Optuna search over 79 real-hardware trials.

Main Results:

• A direct transfer of the baseline controller showed degraded performance due to adaptation-state growth and command saturation.
• The enhanced implementation reduced mean absolute error from 10.476° to 1.054°.
• Internal command saturation ratio decreased from 0.450 to 0.012.
• Achieved significant improvements within the actuator's safety envelope over 300 seconds of operation.

Conclusions:

• A reproducible methodology for deploying and enhancing Nussbaum-based PID control on low-cost manipulators was established.
• Effective controller deployment requires careful consideration of adaptation management, actuation mapping, and hardware-aware optimization.
• The enhanced controller demonstrates robust performance improvements in real-world robotic applications.